At the time of speaking by telephone, speech recognition, speech recording, and the like, it is preferable to collect only target speech (the voice of a speaker). However, in some cases, a sound other than a target speech such as a background noise exists depending on the usage environment of a speech input device. Therefore, the development of a speech input device having a function which enables to accurately extract a target speech, i.e., which cancels the noise even in a case where the device is used in an environment with a noise, has been advanced.
Further, in recent years, the downsizing of electronics has been advanced, and a technology for downsizing a speech input device has become important.    Patent Document 1: JP-A-2007-81614